Cretaceous Software Engineering – Why I don't use AI to write code (alexsuzuki.com)

A self-described "Cretaceous" software engineer explains why they largely avoid using generative AI to write code, arguing that coding by hand is itself a form of thinking that builds an indispensable mental model. Because most of their work is maintenance and bug-fixing (only ~20% new features), that internal understanding lets them localize issues to a few lines and reason about extensions quickly — a payoff they see as worth the slower initial coding. They’ve experimented with tools (e.g., JetBrains Junie) and occasionally use LLMs for one-off tasks, shell scripts, or scaffolding unfamiliar formats like CI YAML, but find coding agents unsatisfying and the generative part of AI unexciting. Non-generative AI uses, such as vector-embedding search for video in an art project, are called out as genuinely useful. The piece is significant because it highlights real-world trade-offs: productivity versus craft, short-term speed versus long-term maintainability, and the cultural impact of automating creative/engineering work. The author worries about devaluing skills (hesitating to teach their child to code), the appropriation of IP on the web, and new workflows where people merely review agent-generated PRs. Ultimately this is a values-driven stance — enjoying hands-on building (ikigai) — and a pragmatic reminder that AI is a tool with nuanced implications for how we learn, debug, and preserve software expertise.